Introduction
Novel therapeutic options are sorely needed to target glioblastoma (GBM), a notoriously treatment-resistant brain cancer. GBM is a leading cause of cancer-related death in the pediatric and adult populations, with most patients succumbing within 1-2 years[
1,
2]. The standard therapies are inadequate, and their toxicities lead to severe life-long morbidity in the small number of patients that survive[
2]. Despite this grim prognosis, GBM is an orphan disease that is in general not a priority for new drug development[
3‐
5]. Moreover, the biology of GBM is complex and much remains to be learned about the putative key signaling pathways before they can be therapeutically exploited[
6]. In view of the unmet and urgent clinical need, we were motivated to pursue recent data indicating that GBM may respond to some FDA-approved agents not previously identified as GBM therapeutics[
7,
8]. The
in vitro screening of a broad range of drugs already approved for other indications is attractive as
in vivo toxicity and pharmacology are well defined, and such compounds can enter GBM clinical trials rapidly either as single agents or as combinations.
Accordingly, our goals were to identify and characterize single and combination agents having anti-GBM activity that we can potentially introduce into clinical trials quickly. To this end, using GBM cell lines and patient-derived GBM cell cultures, we screened a 446-compound NIH Clinical Collection (NCC) library comprising FDA-approved drugs. To further improve the anti-GBM potency of these drugs, we tested various drug combinations and compared them to single drug treatment.
Our screening strategy included multiple human GBM cell lines of different origins in order to provide cross-validation of findings. These cell lines included 4 established serum-grown, immortalized human GBM lines, 4 patient-derived stem cell like GBM primary cells grown as neurospheres, and 2 primary patient-derived GBM lines grown as adherent cultures[
9‐
14].
We did not confine our screening to only adherent GBM stem cell lines despite reports claiming that such lines remain undifferentiated longer and constitute a simpler, less variable assay[
11,
14]. It is not yet firmly established that the major therapeutic target in GBM is just the cancer stem cell tumor compartment and there are indications that other cell types within GBM may assume stem cell characteristics through genetic or epigenetic events[
14‐
16]. In contrast to a single type or lineage of cells, neurospheres contain a mix of GBM stem cells and differentiated cells, which is more reflective of the composition of human GBM tumors[
17]. Further, when dissociated neurospheres are implanted orthotopically, they are highly tumorigenic and authentically recapitulate the invasive natural history, composition, and histology of GBMs growing in humans[
16]. Hence we report the identification of NCC active compounds through our screening approach on patient-derived stem cell-like GBM primary cells.
Our initial screening identified 22 compounds active against GBM (>50% cell death)[
18] at pharmacological doses. These 22 compounds encompassed 11 drug classes. In particular, we found that the statin, pitavastatin, effectively induced cellular autophagy and suppressed tumor cell MDR-1 protein, to impressively enhance the potency of irinotecan, a topoisomerase 1 inhibitor used in cancer treatment[
19‐
21]. This work newly identifies FDA approved drugs and drug combinations, notably pitavastatin and irinotecan, that may be useful for GBM treatment, and draws attention to the potential value of
in vitro screening of approved compounds not currently used to treat GBM.
Materials and methods
Overview of cell-based (in vitro) screening for potential anti-GBM compounds
We acquired 446 small molecules that completed human clinical trials from the NIH Clinical Collection (NCC). The initial broad screen was performed on U87 cells plated at 2000 cell per well on 96-well plates incubated overnight. All compounds were added to the plates to attain a final concentration of 10 μM. After 72 hours of incubation with drugs, the inhibition of cell proliferation was quantified by the Alamar Blue viability assay. Briefly, after incubation, Alamar Blue (#BUF012B, AbD Serotec) was added directly to the culture medium, and the fluorescence measured at 560/90 to determine the number of viable cells (Infinite M200, Tecan Group Ltd.). The IC50 values were calculated using commercially available software (Prism®, Graphpad Software, La Jolla, CA). We defined active compounds as those eliciting a greater than 50% reduction of cell viability in three independent screens. The 15 most potent and available drugs or compounds were then re-screened with other established glioma cell lines (LN443, A172 and U118), with the four patient derived GBM stem cell like primary neurosphere lines (SK72, SK262, SK429 and SK660), and with 2 GBM stem cell like primary cells (SK72 and SK262) grown as adherent culture. Pitavastatin was also tested in combinations with the other 12 compounds. The IC50 values were determined with and without pitavastatin (2 μM), using the Alamar blue assay as described above.
Isolation, culture, and compound activity testing with patient derived GBM cells
Human GBM samples
Fresh human GBM material was acquired from 4 GBM surgical patients and cultured as previously reported[
9,
13]. Briefly, the dissociated tissue was washed, filtered through a 30 μm mesh and plated onto ultra-low adherence flasks at a concentration of 500,000 to 1,500,000 viable cells/ml. The stem cell isolation medium included human recombinant EGF (20 ng/ml), human bFGF (10 ng/ml) and heparin (2 μg/ml). Sphere cultures were then passaged by dissociation, washed, resuspended in neural stem cell culture medium (#05750,Stemcell Technologies), and plated on ultra low-adherence 96 well plates at 2000 cells per well for all subsequent drug testing.
Alternatively, patient-derived dissociated GBM tissues were plated onto laminin-1 coated plates (Sigma, 3-5 μg/ml). Cell populations were dissociated using Acutase (Sigma) and expanded for 5-10 passages, then plated on flat bottom for drug testing.
Confirmation of stem cell marker expression
Primary neurospheres were cytospun onto glass slides. Adherent primary cultures were grown onto Permanox™ chamber slides (#70380,EMS). Cells were incubated with human Nestin antibody (M1259, R&D system) and then with fluorescein-labeled secondary antibodies, then stained with DAPI. The cells were visualized under a UV microscope (Olympus BX51).
Drug testing and survival assay
As explained above, cells were seeded onto either regular or ultra-low adherence 96 well plates and incubated for 18-24 hours and then treated with vehicle control or single drugs or drug combinations. After 96 hours of incubation, Alamar Blue was added directly to the culture medium, and the fluorescence measured at 560/90 after 4-12 hours to determine the number of viable cells. The IC50 was calculated.
Prediction of blood brain barrier (BBB) permeation by active compounds
Although ample evidence has demonstrated that drugs of virtually any size or chemotype can enter brain tumor via leaky tumor microvessels, the ability to penetrate the intact blood brain barrier (BBB) is reasonably hypothesized to be useful for treating tumor cells infiltrating normal brain tissue along fiber tracts[
22‐
25]. Hence we estimated the capacity of active anti-GBM compounds to cross the BBB. We used standard software to calculate the Log BB value: Log BB = -0.0148 PSA + 0.152 CLogP + 0.139; PSA = polar surface area, p = octanol/water partition coefficient[
26,
27].
Determination of cell cycle, autophagy, and apoptosis
Cell cycle analysis
GBM cells were seeded into 10 cm dishes at a density of 1 × 106, cultured overnight followed by the addition of 3 μM pitavastatin with 24 or 48 hours of incubation. Cells were trypsinized and fixed in 70% ethanol for 30 minutes, incubated with 25 μg/ml propidium iodide (PI) and 250 μg/ml RNAase in PBS for 1 hour at 50°C. After PI staining, cells were analyzed via flow cytometry (Canto, BD FACS), and the percentage of cells in G0/G1, S and G2 phases were calculated by ModFit LT software version 3.0.
Detection of caspase activity
Caspase-3 activity was measured with the Invitrogen Enzcheck caspase-3 assay kit #2 according to the manufacture’s protocol. Briefly, 3 × 106 U118 cell were cultured and pitavastatin, irinotecan or the combination was added to the medium for 12 or 24 hours. Then 106 cells were lysed, DEVD-R110 fluorescence substrate was added, and the fluorescence signal was measured and compared with a standard curve (Infinite M200, Tecan). Caspase 3/7 activity was measured by the Apo-One caspase3/7 Kit (#G7790, Promega). 20,000 cells were seeded on to 24 well plates, pitavastatin and vehicle were added, followed by incubation (8, 12, 24 and 48 hours) and caspase 3/7 activity was measured using a fluorescence-based substrate.
Detection of autophagy markers-GFPLC3 punctuation
Retrovirus carrying the GFPLC3 was produced by transfecting the 293GP2 cells with the pVSV-G and pBABE-puro-GFPLC3 plasmids (GFP fused with LC3 gene at the N-terminal, Addgene #22405). Retroviral supernatants were harvested 48 hours later. U87, U118, U251 cells were seeded at a density of 2 × 105 in 6 well plates and infected 24 hr later with the VSV-G/GFPLC3 virus. Stable cell lines were selected for 1 week in 1 μg/ml puromycin. GFPLC3-expressing lines were seeded onto 24 well plates and treated with 1 μM pitavastatin for 48 hours. Presence of GFPLC3 punctuation, which is a marker of autophagy was detected by UV microscopy (Olympus IX81).
Western blot analysis for autophagy, apoptosis, and multidrug resistance protein
LC3 (autophagy), caspase-3 (apoptosis), and MDR-1 (multidrug resistance protein 1 or P-glycoprotein 1,ABCB1) and tubulin (as loading control) were detected by western blotting following drug treatment. Cell lysates (10-50 μg protein) were loaded on to either 14% SDS-PAGE gel (for LC3) or 4-12% gel (for caspase-3, and MDR-1), proteins transferred to PVDF membrane and probed with primary antibodies (L8918 Sigma, 9662 Cell signaling, Sc-8313 Santa Cruz, T4026 Sigma). The resultant protein bands were visualized by a supersignal kit (#1856136,Thermo Scientific) after incubation with HRP-labeled secondary antibodies.
Multi-drug resistance assay
A cell-based fluorescence assay kit (Cayman Chemical Company, MI, USA, #600370) was used to evaluate modulation of the MDR-1 protein by drugs. Calcein AM is a hydrophobic non-fluorescent dye that easily permeates living cells. The hydrolysis of Calcein AM by intracellular esterases produces calcein, a hydrophilic strongly fluorescent compound which is retained in the cell cytoplasm and can be measured using excitation and emission wavelengths at 485 nm and 535 nm, respectively. Calcein AM is a substrate of MDR-1 protein P-gp, which causes its rapid extrusion from the plasma membrane, preventing accumulation of the fluorescent calcein inside the cytoplasm. Therefore measurement of fluorescent calcein allows for detection of MDR activity in live cells. Hoechst Dye staining of nuclei measured using of excitation and emission wavelengths 355 nm and 465 nm respectively to normalize cell numbers in well. GBM cells were seeded at 5 × 104/well overnight, then pitavastatin was added to final concentration of 1, 3 and 10 μM. Twenty four hours after treatment, cells were incubated for Calcein AM/Hoechst Dye solution for 15 min, then fluorescent Calcein retention was measured 20 μM Verapamil or cyclosporine A treatments for 20-30 min as positive control of MDR-1 inhibition followed as the manufacturer’s protocol. The results were expressed as ratio of Calcein AM/ Hoechst signal (C/H ratio). Photomicrographs were taken using fluorescence microscopy.
GBM patients’ survival and free disease status relative to MDR-1 (ABCB1) expression
The GBM patient data were obtained from The Cancer Genome Atlas (TCGA) public data portal, and analyzed using the cBio Cancer Genomics Portal. This system is developed and maintained by the computational biology center of Memorial Sloan-Kettering Cancer Center. We investigated and regrouped GBM patients according their MDR-1 (ABCB1) expression. Firstly, we required the patients/case ID with the MDR-1 expression in all TCGA GBM provisional databases. The mRNA expression z-scores threshold were set as ±1 in our analysis. Then, we regrouped all the patients into 3 groups according to their MDR-1 expression as up-regulated, normal, and down-regulated. Finally, we inputted the down-regulated or up-regulated patients ID with normal expressed patients to select patient/case set to analyze patient survival and free disease status data. The Kaplan-Meier curves were drawn based on these analyses.
Animal studies
The in vivo studies were performed on nude mice to evaluate the drug effects on inhibition of tumor growth. 2 × 106 U87 cells were subcutaneously transplanted into the right and left flanks. Initial tumor growth was monitored every 3 days. Drug administration was initiated when the tumors reached a size of 100-120 mm3. Mice were regrouped into 5 groups of 6 mice each, without significant difference in tumor volume before drug treatment. The mice were treated with either PBS as control, low dose of pitavastatin (0.5 mg/kg), low dose of irinotecan (0.5 mg/kg), a combination of pitavastatin (0.5 mg/kg) and irinotecan (0.5 mg/kg), or high dose of irinotecan (5 mg/kg). All drugs were injected i.p. in 200 μl of PBS, once per day, on a 5-days-on, 2-days-off schedule. Tumors size and mice weight were measured 2 times per week. All mice were sacrificed after tumor sizes reach over 1 cm in diameter in the control group. Tumor volumes were calculated as (length*width*width/2). After sacrifice, all tumors were disserted and weighted. The animal protocol was approved by UCSD Institutional Animal Care and Use Committee (IACUC).
Statistical analysis
Activity against GBM cells was assessed by dividing the average number of viable cells (from three replicates per dose) by the average of three controls. At a type I error rate of 0.05, using a one-sided t-test, we calculated 80% power to evaluate whether a decrease in mean percent viable cells was significantly lower than 100%, if the observed mean percentage was 91.4%; we conservatively assumed the standard deviation of the percent viable cells was 15%. For significant difference by t-test (P < 0.05), labeled * at the bar graphs.
To quantify the synergism of drug combinations, the drug combination index (CI) was calculated as described by Chou[
28]. ED
50, ED
75 and ED
90 were defined as the drug dose able to inhibit cell growth 50%, 75% and 90%, respectively, for pitavastatin alone, irinotecan alone and mixture of two drugs (various ratios). A CI < 1 indicates synergy between the two drugs.
Discussion
In the present study, we sought to screen a library of FDA-approved compounds to rapidly identify new, non-GBM drugs that could be readily introduced into GBM clinical trials. Using a platform that employed a wide range of human GBM lines, including clinically relevant patient-derived primary GBM lines, our screening uncovered 22 compounds from different classes with anti-neoplastic activity in GBM. Among others, the cardiovascular drugs statins showed high efficacy in reducing tumor growth both in vitro and in vivo, drawing our attention to these relatively non-toxic cholesterol lowering drugs. The present study demonstrates the potency of pitavastatin relative to other statins. Importantly, our results demonstrated that co-administration of pitavastatin with low-dose chemotherapy, greatly increased the potency of the latter, lowering the IC50 values for irinotecan by 40- to 70-fold, with few adverse effects. Experimentally, we found that statins independently induced autophagy in GBM and that statins may potentiate chemotherapeutic agents by inhibiting MDR-1 function. This was consistent with in silico screening results using our virtual tumor cell technology, which suggested that pitavastatin affects cell viability by inducing autophagy.
Cholesterol has a key role in cell membranes, cell metabolism, cell signaling and has been implicated in tumor development and progression. Therefore, as cholesterol-lowering agents, questions about the anti-tumor effects of statins have been already posed[
42,
43]. Statins decrease cholesterol levels by inhibiting the enzyme HMG-CoA reductase in the liver. In addition, mevalonate, and isoprenoid intermediates such as geranylgeranylpyrophosphate (GGPP) and farnesylpyrophosphate (FPP) in the cholesterol synthesis pathway are also depleted after statin treatment[
44]. Another intermediate, dolichol, an essential substrate for protein N-glycosylation, is also blocked by statins[
45,
46]. Considering that GBMs are highly proliferative taking up large quantities of cholesterol, potentially they may be vulnerable to statin treatment[
47,
48]. However, the mechanism of sensitivity of GBM to statins has not been elucidated. Recent studies have shown that statins may have an anti-GBM effect in xenograft mouse models, by targeting the low-density lipoprotein receptor (LDLR), inducing apoptosis via ERK/AKT pathway[
20,
47]. Other data hypothesize that statins may inhibit tumor growth by inducing autophagy via the NF-κB pathway in human colon cancer cell line[
49]. Our data obtained in both stable cell lines and primary patient samples clearly demonstrated that pitavastatin induced macro-autophagy in GBM cells[
20,
21]. Further experiments are now ongoing to investigate the signaling pathway(s) involved in this effect.
Importantly, we have shown that pitavastatin potentiated the anti-tumor effects of low-dose irinotecan, a topoisomerase inhibitor. Pitavastatin is know to be a substrate of the multi-drug resistance protein, MDR-1, which is overexpressed in GBM upon drug treatment and is partly responsible for the resistance of GBM to chemotherapy. Our data indicate that, in combination with irinotecan, pitavastatin suppressed glycosylation of MDR-1, thereby inhibiting its function and allowing irinotecan to accumulate intracellularly[
37,
39,
40]. Accumulation of irinotecan is likely responsible for the increased apoptosis in the presence of pitavastatin. The MDR-1 expression in cancer cells can be a significant obstacle to the success of chemotherapy. Many MDR-1 inhibitors have been extensively tested in clinical trials but the results have been inconclusive. According to TCGA data, down-regulated ABCB1 (MDR-1) predicted better survival of GBM patients. Combining a statin with a chemotherapeutic agent represents a powerful, potential strategy for circumventing resistance and significantly enhancing efficacy. Here we have confirmed that pitavastatin may improve the therapeutic response to TOPO-1 inhibitors, by inhibiting MDR-1 function, and may be beneficial for GBM patients. It remains to be determined whether other statins exert a similar or a different anti-neoplastic mechanism as compared to pitavastatin, and whether different subtypes of GBM have different sensitivity to pitavastatin or display other mechanisms for statin actions. GBM is a complex and heterogeneous disease that likely accounts for the different results obtained across various studies.
Irinotecan is broadly used in solid cancer therapy, especially in combination with other drugs[
50,
51]. In clinical use, the toxicity of irinotecan is generally manageable and reversible[
52,
53]. However, in some patients it may lead to severe side effects, such as diarrhea and neutropenia that can be life threatening. In our animal model, co-administration of pitavastatin allowed for a reduced dosage of irinotecan and avoided drug toxicity at higher dosage. These data indicate a new strategy to develop better irinotecan-based drug combination.
Based on the promising results of our present study, we are now undertaking additional preclinical studies of GBM to optimize dosing and characterize efficacy, thus providing a solid basis for a clinical trial with pitavastatin and irinotecan for the treatment of glioblastoma patients.
Acknowledgements
The authors wish to thank Gabi Gomez for technical assistance in the laboratory.
Financial support
Grants from NIH (NIH 3P30CA023100-25S8), National Brain Tumor Society David Cook Chair of Research and from Barbara and Joseph Ajello trust fund, Tuttleman Family Foundation and Boston Fire Department/Kenney Foundation to S. Kesari.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
PFJ and SK designed the study and submitted paper. PFJ performed all experiments and collected all data. TA and SV performed in silico analysis. VF, SP, XC help analysis in vitro data. MG collected patients’ samples. RM, IBS, YC, NN helped on collected in vivo data. MM wrote the first draft of paper. SCP and SK revised the manuscript. All authors read and approved the final manuscript.